In this study, we propose to collect survey, real-time location, and ecological momentary assessment data over a two year period in a sample of 300 urban adolescents. This unique data set will then be applied to developing multilevel and actor-based social network models of the co-evolution of substance use behaviors, peer affiliations, and the use and meaning of geographical space over time. We propose a highly contextually specific research approach to ground social networks within the social environment of adolescents' lives. We will use Ecological Momentary Assessment (EMA) methodology via mobile messaging technology to simultaneously assess multiple influences on adolescent substance use in real time. Along with sampled specific coordinate data of location and a series of standard surveys, this approach will integrate the personal, social, and environmental processes associated with initiation and escalation of substance use. The goal of this study is to model the evolution of multi-level mechanisms affecting substance use for urban youth. The design alluded to thus far (and elaborated further below) gives rise to two related data models; (1) a hierarchical three-level longitudinal design (Raudenbush & Bryk, 2002), with individual change in substance use nested within personal networks, which are in turn nested within neighborhoods, and also (2) a non-hierarchical design with individuals non-uniquely linked to locations. The latter is best thought of as a bipartite (or 2-mode) social network (Wasserman & Faust, 1994; Robins & Alexander, 2004), i.e., a network with two distinct types of vertices, in this case locations and individuals, where ties are only allowed between vertex types. Aim 1. Model changes in substance use over 2 years, focusing on the moderating effects of individual social network quality (risk/protection) on neighborhood-level predictors (concentrated disadvantage: low education and employment, high public assistance; drug related crime; alcohol availability).Aim 2. Model changes in substance use over 2 years, focusing on the mediating effects of individual network quality (riskiness) on individual-level predictors: affective, cognitive and behavioral influences. Aim 3. Model change over two years in the co-evolution of substance use and the use of space, defined by a bipartite social network linking individuals to locations, particularly focusing on: (a) Structural tendencies driven by common influence and selection-related mechanisms (e.g. bipartite-graph equivalents of transitivity, reciprocity and other network closure effects typically found in ordinary (1-mode) networks), (b) Main and moderating effects of activity spaces (individuals' interpretation of locations), and (c) Main and moderating effects of neighborhood characteristics (which are properties of locations).